深度学习 python pytorch 报错处理：RuntimeError: Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor) should...

报错处理：RuntimeError: Input type torch.FloatTensor and weight type -torch.cuda.FloatTensor should...

1. 错误名称2. 错误原因3. 修复方法4. mnist数据集测试的案例参考资料

1. 错误名称

return F.conv2d(input, weight, bias, self.stride,

RuntimeError: Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor) should be the same

2. 错误原因

根据stackoverflow的问答，这个错误产生的原因是： You get this error because your model is on the GPU, but your data is on the CPU. So, you need to send your input tensors to the GPU. 意思是输入数据和模型不在一个地方，模型在GPU上，数据在CPU上，应该把数据输入到GPU上面。

3. 修复方法

Stackoverflow给出了几种建议方法：

第一种是添加代码把数据输入到GPU

# You get this error because your model is on the GPU, but your data is on the CPU. So, you need to send your input tensors to the GPU.

inputs, labels = data # this is what you had

inputs, labels = inputs.cuda(), labels.cuda() # add this line

或者这种，把数据tensor和标签修改为模型所输入的位置

x = x.to(device, dtype=torch.float32)

y = y.to(device, dtype=torch.float32)

亲测可用的是下面这个:

model.to(dev)

data = data.to(dev)

如果数据的不会修改，建议model.to(dev)去掉，但是这样的话，测试了似乎是优先选择cpu模式了

4. mnist数据集测试的案例

其实，现在问题症结已经找到了，就是因为用的pytorch的DataLoader，与前面这些不一致，需要通过iter函数遍历数据，所以需要每次都手动把数据喂给GPU

# mnist数据加载

transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize(mean=(0.5,), std=(0.5,))])

train_set = datasets.MNIST(root='./data', train=True, transform=transform, download=True)

test_set = datasets.MNIST(root='./data', train=False, transform=transform, download=True)

batch_size = 32

num_workers = 1

train_loader = data.DataLoader(train_set, batch_size=batch_size, shuffle=True, num_workers=num_workers)

test_loader = data.DataLoader(test_set, batch_size=batch_size, shuffle=False, num_workers=num_workers)

loaders = {'train':train_loader,

'test':test_loader

}

#############################################################################################################

# 模型训练

from torch.autograd import Variable

num_epochs = 10

def train(num_epochs, cnn, loaders):

cnn.train()

# Train the model

total_step = len(loaders['train'])

for epoch in range(num_epochs):

for i, (images, labels) in enumerate(train_loader):

images, labels = images.to(device), labels.to(device)

# gives batch data, normalize x when iterate train_loader

# print(f"images.shape:{images.shape};bx shape:{Variable(images).shape}; by shape:{Variable(labels).shape}; output shape:{model(Variable(images) ).shape}")

b_x = Variable(images) # batch x

b_y = Variable(labels) # batch y

output = model(b_x)

loss = loss_func(output, b_y)

# clear gradients for this training step

optimizer.zero_grad()

# backpropagation, compute gradients

loss.backward()

# apply gradients

optimizer.step()

if (i + 1) % 100 == 0:

print('Epoch [{}/{}], Step [{}/{}], Loss: {:.4f}'

.format(epoch + 1, num_epochs, i + 1, total_step, loss.item()))

pass

pass

pass

用gpu的测试结果

Epoch [1/10], Step [100/1875], Loss: 1.1673

Epoch [1/10], Step [200/1875], Loss: 0.4501

Epoch [1/10], Step [300/1875], Loss: 0.6198

Epoch [1/10], Step [400/1875], Loss: 0.3469

Epoch [1/10], Step [500/1875], Loss: 0.3964

Epoch [1/10], Step [600/1875], Loss: 0.4239

Epoch [1/10], Step [700/1875], Loss: 0.4740

Epoch [1/10], Step [800/1875], Loss: 0.3085

Epoch [1/10], Step [900/1875], Loss: 0.6178

Epoch [1/10], Step [1000/1875], Loss: 0.6015

Epoch [1/10], Step [1100/1875], Loss: 0.1161

Epoch [1/10], Step [1200/1875], Loss: 0.2946

Epoch [1/10], Step [1300/1875], Loss: 0.2689

Epoch [1/10], Step [1400/1875], Loss: 0.3746

Epoch [1/10], Step [1500/1875], Loss: 0.2356

Epoch [1/10], Step [1600/1875], Loss: 0.0904

Epoch [1/10], Step [1700/1875], Loss: 0.2226

参考资料

【1】stackoverflow 问答：RuntimeError: Input type (torch.FloatTensor) and…

好文链接